NeuRec: On Nonlinear Transformation for Personalized Ranking

Authors: Shuai Zhang, Lina Yao, Aixin Sun, Sen Wang, Guodong Long, Manqing Dong

IJCAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on four real-world datasets demonstrated their superior performances on personalized ranking task.
Researcher Affiliation Academia 1 University of New South Wales, 2 Nanyang Technological University, 3 Griffith University 4 University of Technology Sydney
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code No The paper states 'We implemented our proposed model based on Tensorflow3' with a footnote linking to the TensorFlow website, but it does not provide an explicit link to its own source code repository or state that its code is open source.
Open Datasets Yes We conduct experiments on four real-world datasets: Movielens Het Rec, Movielens 1M, Film Trust and Frappe. The two Movielens datasets1 are collected by Group Lens research[Harper and Konstan, 2015]. Movielens Het Rec is released in Het Rec 20112. Film Trust is crawled from a movie sharing and rating website by Guo et al. [Guo et al., 2013]. Frappe [Baltrunas et al., 2015] is an Android application recommendation dataset.
Dataset Splits No The paper states, 'We use 80% user-item pairs as training data and hold out 20% as the test set, and estimate the performance based on five random train-test splits.' It also mentions 'We do grid search to determine the hyper-parameters.' While hyperparameter tuning implies a validation process, a specific validation split percentage is not explicitly stated.
Hardware Specification Yes We implemented our proposed model based on Tensorflow3 and tested it on a NVIDIA TITAN X Pascal GPU.
Software Dependencies No The paper mentions 'implemented our proposed model based on Tensorflow3' but does not specify a version number for TensorFlow or any other software dependencies.
Experiment Setup Yes For all the datasets, we implement a five hidden layers neural network with constant structure for the neural network part of Neu Rec and use sigmoid as the activation function. For ML-Het Rec, we set the neuron number of each layer to 300, latent factor dimension k to 50 and dropout rate to 0.03; For ML-1M, neuron number is set to 300, k is set to 50, and dropout rate is set to 0.03. The neuron size for Film Trust is set to 150 and k is set to 40. We do not use dropout for this dataset; For Frappe, neuron size is set to 300, k is set to 50 and dropout rate is set to 0.03. We set the learning rate to 1e 4 for ML-Het Rec, ML-1M and Frappe. The learning rate for Film Trust is 5e 5. For ML-Het Rec, ML-1M and Film Trust, we set the regularization rate to 0.1, and that for Frappe is set to 0.01.